Lyra Evidence Dashboard

Atomic factual statements extracted from literature by LLM. Each claim represents a single verifiable assertion about the research hypothesis. Claims are the fundamental unit of evidence analysis.

Claims with ≥1 "refutes" evidence edge from NLI classification. Indicates areas of scientific disagreement.
Controversy Score = refute_count / (support_count + refute_count + 1). Range: 0–1.

Posterior probability from Bayesian updating. Prior=0.50 (neutral). Each NLI evidence edge updates the belief.
>0.50 = supports hypothesis | <0.50 = refutes | =0.50 = neutral.

Combined metric of source importance: claims_supported × (1 + log(citations + 1)). Higher scores indicate sources that provide more evidence for the hypothesis. Displayed in Key Supporting Sources.